A deterministic diagnostic logic tree
for a probabilistic ranking system.
Every site that fails to rank has one binding constraint
Bottleneck removal fixes most SEO performance problems. A site that fails to rank has a binding constraint, one layer that it cannot pass. Work downstream of that constraint does nothing until the constraint clears.
The clinical method gives three instructions. Test the layers in order. Stop at the first binding constraint. Put the capital into that constraint before you go downstream. The method prevents one expensive error: a team optimizes Layers 4 through 7 while the site carries an open Layer 2 access problem.
One distinction makes the method work. Layer 1 sets the ceiling. A penalized domain cannot rank for the query class at any price. Layer 7 sets the weight. An eligible domain can rank, but it must outweigh the incumbents. The two problems look the same in a rank tracker, and they take opposite interventions.
The output of the framework stays probabilistic. The diagnosis does not. A layer is not an opinion. A layer is a testable and falsifiable condition. The content passes it, or the content fails it.
The investment screen runs before Layer 1
A binding constraint is worth finding only when the channel, the page type, and the query justify the work. The investment screen is not a diagnostic layer. The screen is a strategic filter that decides whether to run the diagnostic at all, and against which pages and queries. Each screen must clear before the next one runs.
Channel Qualification
Is organic search the right acquisition channel for this business?
Evaluate
Is organic search a meaningful acquisition channel for this business model? Some B2B businesses and service models capture demand through referral, through directories, or through paid media. Organic search is not always the right investment.
Failure Signature
A business where organic search is structurally not the right acquisition channel. If this screen fails, neither page type selection nor query-level analysis matters. Stop here.
Page Category Allocation
Do the page types match the structure of demand in this vertical?
Evaluate
- —Business model classification: local service, e-commerce, B2B SaaS, publisher. Each has a distinct organic acquisition surface and page type requirement.
- —Page type alignment: do the proposed content investments match the structure of demand in this vertical?
Failure Signature
A local service business builds informational content clusters instead of location-service pages. The page type is wrong for the business model, and no amount of optimization corrects that. A wrong page type at Screen 1 makes the query-level screen irrelevant.
Query-Level Expected Value
Does the expected return justify the investment required?
Evaluate (in sequence)
- 1.Realistic rank ceiling. Given current domain authority, SERP composition, and incumbent topical authority, what is the best achievable position? If the gap to incumbents is large, the ceiling may be 6-8, not 1-3.
- 2.SERP structure risk. Before you estimate a return, describe what an organic position yields on this SERP. Is the organic click floor acceptable in the best case position? Some SERPs suppress organic clicks by structure. Ads, AI Overviews, dominant aggregators, the maps pack, and the feature units together cut the organic share until position 1 captures marginal traffic. Other SERPs stay clean. The judgment is categorical, not numeric. Does this SERP give organic clicks, or does the investment case depend on generative inclusion and brand visibility?
- 3.Generative surface potential. Can this query class win an AI Overview or an LLM citation? Generative inclusion changes the investment case when the SERP suppresses traditional CTR.
- 4.Commercial conversion likelihood. Does the traffic this query class generates convert, or does it attract users with no intent to buy?
- 5.Expected return vs. effort. Combine the above. If everything goes right, does the outcome justify the investment required to get there?
Failure
A query looks achievable, but it does not pay off. The rank ceiling is reachable. The return is not. SERP feature suppression, the authority gap, or low conversion intent keeps the expected value below the cost, even when the page reaches the ranking.
Each layer fails in a different way
The investment screen clears the economics. The eight layers test the mechanics. Each layer states one testable condition and has one failure signature. Run them in order and stop at the first failure.
Eligibility
Clinical Question
“Is this site eligible to rank for this class of query at all?”
What to Look For
- —Domain authority relative to the SERP composition
- —Algorithmic penalty indicators at the site or category level
- —Content quality signals at the domain level (average page quality)
- —YMYL classification risk and associated trust threshold requirements
- —History of manual actions or devaluation patterns
Failure Signature
A team publishes optimized content into a domain that Google penalizes at the query category level. The ceiling is zero, not low. On-page optimization and link building cannot move content past a closed eligibility constraint.
Layer Note
The eligibility ceiling and the competition weight at Layer 7 are different constraints, and they need different interventions. A site can clear Layer 1 and still carry a high competition weight. A consultant who reads a ceiling problem as a weight problem buys links for a penalized domain. That substitution is the most expensive mistake in SEO capital allocation.
Access
Clinical Question
“Can the discovery systems (search crawlers and AI agents) reach, render, and index the content?”
What to Look For
- —robots.txt rules for Googlebot, GPTBot, PerplexityBot, and other AI agents
- —JavaScript rendering dependency: does the page return meaningful content to a raw HTTP request?
- —Crawl budget allocation across the site architecture
- —Core Web Vitals and server response time under crawl conditions
- —WAF and bot detection rules inadvertently blocking legitimate crawlers
- —Canonical signals and noindex directives
Failure Signature
The content exists in the browser, but not in the crawl. JavaScript-rendered pages return empty or near-empty bodies to the crawlers. Default WAF rules, written before the AI crawler wave, block the AI agents. A Layer 2 failure looks like a content problem or a competition problem.
Representation
Clinical Question
“Does the structural representation of the content match the subject of the page?”
What to Look For
- —Title tag accuracy: does it reflect the actual primary query intent?
- —H1 hierarchy and heading structure clarity
- —Schema markup correctness, completeness, and appropriate type selection
- —Canonical signals pointing to the correct authoritative URL
- —Entity disambiguation: is the page's primary topic unambiguous?
- —Freshness signal accuracy: does the CMS label evergreen content as news?
Failure Signature
A news CMS stack puts freshness signals on evergreen content. Product pages use Article schema. Category pages carry no schema. Title tags describe the site, not the page. The representation mismatch is the most common cause of 'we have great content but we cannot rank.' The writer wrote the content correctly. The template labels it incorrectly.
Retrieval, Lexical
Clinical Question
“Does the page's vocabulary match the vocabulary of the target queries?”
What to Look For
- —Keyword presence in high-weight positions: title, H1, opening paragraph
- —Query variant and synonym coverage throughout the document
- —Entity surface form alignment with the words searchers use for the concept
- —Natural language density: is the target vocabulary present without stuffing?
Failure Signature
The page uses concept-level language where searchers use product, brand, or colloquial language. The opposite failure also occurs: over-optimized copy repeats the target term at a density that modern models deprioritize. A lexical retrieval failure is an editing problem, not an architecture problem.
Retrieval, Semantic
Clinical Question
“Is the page's topic model sufficient to rank within the relevant semantic cluster?”
What to Look For
- —Entity coverage: are the related entities mentioned, linked, and contextualized?
- —Topical completeness relative to the top-ranking documents
- —Semantic neighbor pages on the same domain: is there a topical cluster?
- —Internal linking structure supporting the semantic relationship between pages
Failure Signature
A thin page answers the head term, but it omits the semantic context that builds topical authority in the index. An orphaned page carries strong on-page signals with no internal link context. The semantic cluster signal is absent. A semantic retrieval failure is a content depth problem and a site architecture problem.
Retrieval, Generative
Clinical Question
“Is this content retrievable and citable by AI-powered answer systems?”
What to Look For
- —Crawl access for AI agents: GPTBot, PerplexityBot, ClaudeBot, and similar
- —Structured answer formatting: direct question-answer pairs, definition blocks, summary sections
- —Citation-friendly sourcing: primary data, original research, attributed claims
- —Content rendering without JavaScript dependency
- —YMYL trust signals for AI grounding: authorship, methodology transparency, sourcing
Failure Signature
Content that wins on the ten blue links can fail in a generative context. The page buries the answer in narrative prose, needs JavaScript to render, or cites secondary sources instead of primary data. Layer 6 is a parallel retrieval track, not a branch of Layers 4 and 5. A page can pass Layers 4 and 5 and still fail Layer 6.
Layer Note
Layer 6 is the entry point for the S.A.G.E. optimization model. When generative retrieval is the binding constraint or the target track, S.A.G.E. (Structured Signals / Authority & Authenticity / Generative Access / Experience Depth) is the prescription. See the S.A.G.E. framework for the specific optimization protocol.
Competition
Clinical Question
“Can this page win the ranking competition for this query at this point in time?”
What to Look For
- —SERP composition: who holds positions 1-3, what content types, what authority levels
- —Authority gap at domain and page level
- —SERP feature concentration: featured snippets, AI Overviews, local pack, image carousels
- —Content differentiation gap: what does the incumbent have that this page lacks?
- —Link profile gap: quantity, quality, and topical relevance of inbound links
Failure Signature
A team pushes equal-quality content against entrenched incumbents with no differentiation strategy. Competition is a weight problem, not a ceiling problem. A site that clears Layer 1 but fails Layer 7 faces a capital allocation question. Build authority first on achievable queries, or differentiate on content depth and on data.
Layer Note
The evidence loop operates at Layer 7. Early wins on achievable queries compound domain authority, and that authority lowers the Layer 7 threshold for harder queries in the same cluster. Follow a fixed sequence. Win Layer 7 on medium-difficulty queries, bank the authority priors, then re-enter the harder queries.
Satisfaction
Clinical Question
“Does the content satisfy the user's complete search intent and generate the engagement signals that reinforce ranking?”
What to Look For
- —Time-on-page and scroll depth patterns across organic sessions
- —Pogo-sticking rate: immediate return to SERP after landing
- —Task completion rate: did the user accomplish what they came for?
- —Return visit rate from organic entry points
- —Conversion rate from organic: the ultimate downstream satisfaction signal
Failure Signature
A page holds a ranking, but it carries persistent pogo-sticking and short session duration. Google reads the satisfaction signals as direct inputs into its quality assessment. A page that ranks on old authority priors while the satisfaction signals fall is not in a stable state. The ranking decays as the negative engagement signal compounds.
The layers run in order, but they do not run in isolation
Layer 8 closes the sequence, but the sequence is a diagnostic priority, not a causal chain. A cleared Layer 2 (access) does not make Layer 3 (representation) work. Each layer has an independent failure mode. The sequence tells you where to look first. The sequence does not tell you where to stop.
Layers 4, 5, and 6 are parallel retrieval tracks, not alternatives. A page can pass all three, or any combination of them. The generative track at Layer 6 needs different optimization from the traditional tracks at Layers 4 and 5. A team that treats the three as one track writes content that wins in one channel and stays invisible in the other.
The Evidence Loop
A site clears Layer 7 quickly on achievable queries. Those early wins compound domain authority over time, and that authority lowers the Layer 7 threshold for harder queries in the same topical cluster.
The loop is not a hack. The loop is how organic authority compounds. A low-authority challenger wins the tail queries and the mid-tier queries first. Those wins build the authority priors in the topic cluster. The challenger then re-enters the head terms from a stronger baseline, with no bought links.
Banksparency.com runs the loop live, at 9,000 to 10,000 monthly pageviews, with no link building and no editorial hours. The architecture solves Layers 1 to 6 by design. Layer 7 is the active frontier. The query sequence builds the authority priors in the banking data cluster. Data granularity is the differentiation strategy for the query windows that a low-authority challenger can win before it reaches authority parity on the head terms.
The coupling has one more consequence. A site that fails Layer 1 (eligibility) for competitive head terms can pass Layer 1 for the long-tail variants in the same category. Run the diagnostic once for each query class, and the actionable segments become visible. The result is not one verdict for the whole domain. The result is specific to the query class.
Five rules decide whether the sequence works
The coupling rules explain how the layers interact. These five rules govern how to run them.
Run layers in order. When Layer 2 (access) holds the constraint, a link campaign destroys capital. When Layer 1 (eligibility) holds it, every downstream task is waste until you clear the eligibility problem. You cannot skip the sequence.
The 60-minute triage is not the full diagnostic. The triage is a pre-check that shows which layers need a deep investigation. A full analysis of Layer 2, Layer 3, and Layer 7 needs dedicated tooling, log file analysis, and comparative SERP research.
Layer 6 is a parallel track, not a branch. Generative retrieval needs its own optimization. A page can rank on the ten blue links and stay invisible in AI Overviews, because the two retrieval mechanisms differ. Treat Layer 6 as a separate investment, not as a side effect of the Layer 4 and Layer 5 work.
The investment screen is not part of the diagnostic. Channel qualification, page type alignment, and query-level expected value are economic decisions, not pass or fail tests. A team that runs the diagnostic before the screen optimizes the right queries for the wrong page type on the wrong channel.
Layer 8 (satisfaction) is not a vanity metric. Google reads pogo-sticking and session duration as active inputs into its quality assessment. A ranking that rests on old authority priors while the satisfaction signals fall is a ranking in decay. Satisfaction is a layer, not an afterthought.
Sixty minutes narrows the search to one layer
The table below compresses the screen and the eight layers into one reference. Run it before you commit to a full diagnostic. If every check clears in sixty minutes, go to Layer 7 and Layer 8. The binding constraint is competition or satisfaction.
| Layer | Name | Clinical Question | Quick Check |
|---|---|---|---|
| Pre-layer | Investment Screen | Right channel? Right page type? Right query? | Channel fit → business model classification → Demand x SERP risk x Difficulty |
| Diagnostic Layers | |||
| Layer 1 | Eligibility | Can we rank at all? | DR gap to SERP median, penalty check, YMYL risk assessment |
| Layer 2 | Access | Can crawlers reach and render it? | Raw HTTP body check, robots.txt audit, crawl coverage report |
| Layer 3 | Representation | Does the page say what it does? | Title / H1 / schema audit, freshness signal audit |
| Layer 4 | Retrieval, Lexical | Does vocabulary match target queries? | Keyword presence in high-weight positions, variant coverage |
| Layer 5 | Retrieval, Semantic | Is the topic model complete? | Entity coverage vs. top-ranking pages, internal link cluster |
| Layer 6 | Retrieval, Generative | Is this AI-retrievable? | AI crawler access, structured answer presence, primary data sourcing |
| Layer 7 | Competition | Can we win the ranking battle? | Authority gap, SERP feature concentration, differentiation gap |
| Layer 8 | Satisfaction | Does the user stay and convert? | Pogo-sticking rate, time-on-page, task completion, return visits |
A deterministic tree finds the constraint that a probabilistic system hides
A ranking stays probabilistic. The diagnosis does not. Each layer states a condition that a site passes or fails, so eight tests turn an unpredictable output into one answerable question. Which layer blocks this query class today? Banksparency answers that question from zero, with no backlinks, no editorial team, and no inherited authority.